Spaces:

RFTSystems
/

RFTs_Forecasts

Running

App Files Files Community

RFTs_Forecasts / README.md

RFTSystems

Update README.md

b6e17ab verified 4 days ago

preview code

raw

history blame contribute delete

4.75 kB

metadata

title: RFTs Forecasts
emoji: 🐢
colorFrom: pink
colorTo: purple
sdk: gradio
sdk_version: 6.1.0
app_file: app.py
pinned: true
short_description: RFTs Weather phenomena live forecasts
license: other
thumbnail: >-
  https://huggingface.co/static-proxy/cdn-uploads.huggingface.co/production/uploads/685edcb04796127b024b4805/6Hp-B0MSDCUoC8dsijWJg.png

Rendered Frame Theory — Live Prediction Console (Open Method)

This Space runs live, transparent, recomputed-from-scratch signals for four domains:

Atmospheric (location-based)
Seismic (region or local-radius, depending on your Seismic Mode)
Magnetic (global)
Solar (global)

No hidden steps. The app surfaces the exact values it computes: z, τ_eff, Ω_obs, α_R, index, and the decision rule that fired.

What changes with location (and what does not)

Location input

Used for Atmospheric (Open-Meteo hourly at the geocoded lat/lon).
Used for Seismic only in “Local radius” mode (USGS events within your chosen km radius).
Not used for Solar (GOES X-ray flux is global).
Not used for Magnetic (Kp index is global).
Not used for Seismic in “Region” mode (region selector controls the filter).

If you type a different city and Solar/Magnetic stay unchanged: expected.

Seismic modes (important)

This Space supports two seismic views:

1) Region mode

Counts earthquakes in the selected region over the last 24 hours with M ≥ 2.5.
This is a regional stress monitor. It is not “near your city”.

2) Local radius mode

Counts earthquakes within your chosen radius (km) around your typed location over the last 24 hours with M ≥ 2.5.
This is still not a time/epicenter predictor — it’s an activity density monitor.

What this is / what this is not

This is

A live multi-domain regime detector that turns raw public feeds into an explicit RFT-style stress coordinate z, maps it to τ_eff, then produces an index, then assigns labels using fixed thresholds.

This is not

A guaranteed “prediction engine”.
A precipitation model, radar nowcast, or full NWP weather model.
An earthquake time + epicenter predictor.
A CME arrival model or flare timing predictor.
A local magnetometer or grid impact model.

When live data is missing or too short, the domain is DISABLED instead of guessed.

Forecast Receipts (durability + verification)

Each run generates a downloadable Forecast Receipt (JSON) to make results inspectable over time.

The receipt includes:

Source URLs + request parameters + timestamps
sha256 hashes of upstream payloads
Computed intermediates and the exact rule_fired
Environment snapshot (versions + constants)

Optional durability mode:

Enable “embed raw upstream payloads” to include raw_b64 inside the receipt for stronger offline verification.

You can later upload the receipt under the Verify Receipt tab to:

Validate receipt structure
Verify embedded payload hashes (if present)
Recompute z / τ_eff / index / label and confirm they match the stored outputs

Instant verification links (official sources)

Use these to falsify the live status immediately:

Open method (equations used in-app)

Shared core:

τ_eff = 1.38 · ln(1 + z)
Ω_obs = 2π / T_earth (T_earth = 365.2422 days)
α_R = 1.02
Index = Ω_obs · τ_eff · α_R

z definitions:

Atmospheric: z_atm = clamp( clamp(ΔT/10,0..2) + clamp(|ΔP|/12,0..1.5), 0..3 )
Seismic: z_seis = clamp( clamp(N/60,0..1.5) + clamp(max(0,Mmax-4)/2.5,0..1.5), 0..3 )
Magnetic: z_mag = clamp( (Kp_last/9) + (drift/2) + 2·|slope|, 0..3 )
Solar: z_solar = clamp( ln(F_mean/1e-8)/10, 0..3 )

Decision thresholds are printed per-domain in the agent output as rule_fired.

Practical note: “Why does my city show X seismic events?”

If Seismic Mode = Region, the count is for the selected region (e.g., EMEA) over the last 24 hours at M ≥ 2.5.
Switch to Local radius mode to make the count location-dependent.